Re: GUID as primary key

Mark Bole schrieb am 20.06.2005 in
<66Bte.353$N22.40_at_newssvr21.news.prodigy.com>:

> Andreas Mosmann wrote:

>> Mark D Powell schrieb am 20.06.2005 in 
>> <1119273534.757735.55110_at_g49g2000cwa.googlegroups.com>:
>> 
>>> Andreas, Billy provided the necessary statement, but what is the point
>>> of adding a GUID to every table?  Does every table really need a GUID?
>> 
>> We hope it will make our live easier. I told Billy, that the application 
>> will work in several departments. Sometimes the connection will be cut 
>> but the Software still has to work. We hope that merging the data when 
>> connection is active again will be easier and that the developement of 
>> the application will be easier because we dont have to decide, whether 
>> connected or not.
>> 

> The OP states "one database", but somehow when a department has lost the
> connection, they are still able to generate data updates, which makes me
> think there is really more than one data store involved (maybe not all
> Oracle-based).

Correct, there are more than 1 data stores involved. But in that case updates normally will not appear, because records are not allowed to be updated, instead of this a new one is generated with the different value.

> Anyway, that guess, plus the original question specifically about GUID,
> leads me to suggest the following link, which describes three ways to
> generate non-overlapping key values from different source databases.

> http://download-west.oracle.com/docs/cd/B10501_01/server.920/a96567/repmaster.htm#41182
>

I will look at this, but if I understand correctly you do not find any further problems in that way, isn't it?

> -Mark Bole

Andreas Mosmann

-- 
wenn email, dann AndreasMosmann <bei> web <punkt> de